Male ddY mice were housed in a 4-, 6-, or 8-arm radial maze apparatus for 6 hr a day and then removed to a normal cage for fasting until the next day's trial. A water bottle was placed at the central platform. The end of each arm ran to the home cage through a guillotine door (G). Food was placed in the home cage. During the housing in the apparatus, one G was opened, and the remaining Gs were shut. The opened G was changed in random order during the housing period of 6 hr. At the beginning of the trial, all Gs were shut. The mouse was placed on the platform and was permitted to choose the arms until it chose all arms. When the mouse chose the last arm, the G was opened to allow stepping into the home cage. The mice housed in these apparatus learned to go to the home cage without entering already chosen arms within 4-6 days. Scopolamine (0.2 and 0.4 mg/kg, i.p.) impaired the maze performance. A 1- to 4-min delay impaired the performance slightly, but did not show any significant effect depending on the delay intervals. These results suggest that the apparatus is a useful and easy method for estimating working memory and the drug effects thereon in mice.